A recent Data Warehousing Institute conference session, which focused on some possible ways to approach a master data management (MDM) initiative, began with a friendly but very real warning.
"If you've got lots of different copies of what should be the same information, but it's not the same, then potentially that is a business problem," said Cliff Longman, chief technology officer of Kalido, a Burlington, Mass.-based MDM software vendor.
In his presentation, Longman said that companies like Unilever PLC and British Petroleum PLC are solving that business problem today with the help of MDM software and related data governance techniques. He cautioned, however, that because it is such a new discipline, the best MDM practices are still not clearly defined.
"The notion of understanding where your data is stored and managing the potential conflicts that arise when you get copies of that data out of line with one another is what MDM is all about," Longman said. "We have seen some staggering benefits from it and we have seen some things fail."
An increasing number of companies have taken notice of MDM recently because of its potential for delivering the benefits of greater data consistency, which can include more accurate reporting and data analysis, a reduction in errors and an enhanced bottom line.
MDM combines middleware with data governance in an effort to solve the problem of getting source data – and all of the copies of that data which tend to multiply throughout an organization – to sync up.
The middleware defines key business objects such as vendors, products, customers and employees and, ideally, becomes the single source for all of an organization's approved data definitions. The data governance aspect of MDM focuses on aligning business and IT in an effort to identify and fix the flawed business processes which lead to inconsistencies.
Making a master data management business case
Before any IT department can get started on MDM, it will need to get budget approval from the business side of the organization. This means that IT workers will need to quantify the value that MDM will add to the company, according to Longman.
He said it's also a good time for business and IT to start getting used to the idea of working more closely together, because that closeness will come in handy when it comes time to implement new data governance policies.
"You have to provide a business case for this and if you do so, use the right terms that the business people understand," he said, adding that those terms include return on investment (ROI) and total cost of ownership (TCO).
IT managers unfamiliar with ROI, TCO and related concepts should "get a business person to help make the business case," he said.
Those making the business case for MDM will also want to demonstrate that the company has – or can get -- the skills and resources necessary to make the project a success. Companies that have already completed a data warehousing initiative may have an easier time jumping this hurdle, Longman said.
"The skills, principals and techniques that are used to build data warehouses can be used to very good effect in the MDM arena." Longman said. "In actual effect, most data warehouses have at least solved the first part of the problem, which is getting to see what's out there, because you have brought it all into one place."
Making a business case also means getting a handle on exactly what types of data will be covered by the MDM initiative. Longman said the answers to this question tend to vary by company.
"One of the decisions that you make when you embark on an MDM program is to decide the scope of the program and where you start first," he said. "The biggest bang for the buck [may come from] getting your product catalog sorted or getting your customer data shipshape or getting pricing organized -- whatever it is in your company that is most important."
Two approaches to master data management
When Unilever started implementing MDM technology they decided to leave their master data and the copies of that data alone and tackle the data governance issue first, Longman said.
They began by picking some employees and charging them with the responsibility of managing the values of data within operational systems. They then created a process for replicating the master data so that copies are updated when changes are made.
"[Unilever] had set aside an SAP system specifically for holding a copy of the master data, and assigned responsibility to people to manage the data in that system," Longman said. "They used SAP replication technology to manage all of the other copies that were around the organization."
Longman said that Unilever used enterprise application integration (EAI) tools to manage changes and replication on an "as-it-happens" basis.
"As something changes in one system, a message is sent through a message bus and it's replicated to all of the places that that data is maintained," he explained.
Longman added that this type of approach is often more successful at smaller companies than large ones, because it tends to be very expensive.
A second possible approach to implementing MDM technology begins with data harmonization and consolidation, rather than data governance.
Under this approach, Longman said a company would start by dumping all of the chosen data "in all it's gory detail" into one place, such as a data warehouse. From there, the company would begin addressing the data governance issues.
The CTO said this method is particularly useful to companies that want to address analytical issues.
"It's much easier to take data and push it into a data warehouse than it is to take data and push it into operational systems," Longman said. "This harmonization and consolidation stage has to do with getting sight of what you've actually got and moving it into a better position from a data perspective."